Numerical Model-driven Rolling Bearing Fault Diagnosis And Performance Degradation Assessment

Rolling bearings as one of the important parts of rotating machinery,its service status is related to the safety of equipment and personnel.This paper takes rolling bearings as the research object,and explores the methods to improve the accuracy of bearing fault diagnosis and characterize the performance degradation trend under non-smooth operating conditions from four perspectives,such as the fault impulse mechanism,redundant information elimination in variable operating conditions,simultaneous elimination of redundant information in source and target domains under variable operating conditions,and performance degradation assessment under non-smooth operating conditions,in response to the key problems of missing fault samples and large variability of sample characteristics distribution of the same fault type under complex operating conditions.The main research of this paper is as follows:(1)Taking cylindrical roller bearing as the research object,the actual working condition of the bearing is mapped to the dynamic model,and the degree of rolling body slippage,the trend of maximum shear stress and principal stress in the outer ring and the direction of crack expansion are analyzed.The results show that: with the expansion of the fault size,the degree of rolling body slippage becomes more serious;The maximum shear stress generated by the exit fault is higher than that of the entry fault,so the side of the rolling-out fault is prone to crack expansion,and the crack expansion rate increases with the increase of the fault degree.(2)Bearings are usually in non-smooth working conditions,and the distribution of sample features under different working conditions is not consistent.Based on the above situation,NAP-based fault diagnosis method for variable operating conditions of bearings is proposed.Firstly,the constructed dynamics model is used to simulate multiple working conditions and multiple fault types of samples,and the simulation projection matrix of different fault types is obtained by NAP;secondly,the load and speed redundancy information of the measured signal feature samples are eliminated synchronously by using the simulation projection matrix;finally,the bearing variable working condition fault diagnosis is combined with SVM.The discrete variable load,variable speed and continuous variable speed bearing data sets are analyzed.The test results show that the simulation projection matrix effectively eliminates the redundant working condition information.(3)The complex application environment and the extreme lack of fault samples pose difficulties for bearing health status detection.Based on c DNAP,a number-model driven crossdomain variable condition fault diagnosis method is proposed.The simulation data sets are firstly constructed under multiple fault types to solve the problem of missing fault samples.Then the simulation data sets are expanded by WGAN.Finally,based on the simulation sample sets,the cross-domain simulation projection matrices is obtained by c DNAP to align the sample features of variable conditions in the source and target domains,so as to achieve cross-domain variable condition fault diagnosis.The results of the variable condition tests show that the diagnostic accuracy reaches 99% under a very small number of fault samples.Compared with DANN,DSAN and DAAN methods,the proposed method has better performance in diagnosis.(4)To address the problem that the conventional time-domain statistical features are insensitive to weak failures of rolling bearings under non-smooth operating conditions,a performance degradation evaluation index,which characterize the performance degradation trend,is proposed.The sensitivity of statistical characteristics to radial load,speed fluctuation and failure degree is investigated based on the explicit dynamics model of the bearing,and the best reconstructed characteristics that are insensitive to operating conditions but sensitive to the failure degree are obtained.The bearing performance degradation trend was characterized by calculating the FSI of the best reconfiguration characteristics of two adjacent samples.The FSI of the full-life fatigue test data of the bearing under non-smooth operating conditions is calculated,and the results show that the FSI is sensitive to the weak failure of the bearing under non-smooth operating conditions,and is in good agreement with its performance degradation trend.